Generating Synthetic Tabular Data that is Differentially Private

Explore approaches to applying differential privacy in generating synthetic tabular data in this 27-minute talk by Lipika Ramaswamy, Senior Applied Scientist at Gretel AI. Learn how generative models can produce synthetic datasets that preserve statistical qualities without identifying specific records, and understand the importance of mathematical privacy guarantees. Discover a method that combines measuring low-dimensional distributions with learning graphical model representations to create high-quality, differentially private synthetic data. Gain insights into how differential privacy defends against future privacy attacks by introducing calibrated noise into algorithms, providing a robust solution to data privacy challenges in synthetic data generation.

Generating Synthetic Tabular Data that is Differentially Private